Full Stack Developer for Legal Tech

We're building the agentic layer for legal work systems where multiple AI agents draft, review, and reason over contracts and case files with a human in the loop. We're a small team shipping fast, and we need someone who can own features end to end without waiting to be told what to do. What you'll do Build and ship full-stack features (React/TypeScript front to Python or Node back) that put multi-agent workflows in front of real lawyers. Design agent orchestration tool calling, retrieval, evaluation, retries, and the unglamorous plumbing that makes agents reliable instead of demo-ware. Wrestle with latency, cost, and correctness when you're chaining LLM calls in production, not a notebook. Talk to users (yes, actual attorneys) and turn their messy workflows into something that holds up under real document loads. You probably have 4+ years building production web apps, plus hands-on work with LLM/agent frameworks (LangGraph, custom orchestration, whatever you've actually shipped). A real opinion on how to evaluate agent output, because "it looked right in testing" doesn't survive contact with a 90-page MSA. Comfort with ambiguity and a bias toward shipping. Nice to have Legal tech, document-heavy, or other high-stakes/high-precision domain experience. You've felt the pain of hallucination in a context where being wrong has consequences. If you've built agent systems that survived real users, we want to talk.